CN106023169A - Garment-making cutting piece cross stripe alignment method based on image matching - Google Patents

Abstract

A garment-making cutting piece cross stripe alignment method based on image matching disclosed by the invention is implemented according to the following steps: S1, acquiring an image; S2, graying the image; S3, performing threshold segmentation on the image to get a plurality of cutting piece binary images; S4, performing hole filling on each pair of binary images in need of cross stripe alignment to eliminate the jagged edges in each pair of cutting piece binary images; S5, performing AND operation on each pair of cutting piece binary images and the original image to get each pair of segmented cutting piece images; S6, correcting the positive direction of cutting piece of each pair of segmented cutting piece images; S7, selecting a to-be-sewn area of one cutting piece from each pair as a template; S8, matching the to-be-sewn area of the other cutting piece to get a maximum response position; and S9, calculating the moving distance of the cutting piece according to the maximum response position and the original position of the template, and moving the cutting piece according to the moving distance to complete cutting piece cross stripe alignment. Through the method, there is no need for manual cross stripe alignment.

Description

A kind of clothing cut-parts based on images match method to horizontal stripe

Technical field

The invention belongs to clothing cut-parts technical field of weaving, be specifically related to a kind of clothing cut-parts based on images match to horizontal stroke The method of bar.

Background technology

In the clothing link of textile clothing enterprise, workman needs manually by several piece sewing panels together, especially It is for striped, the senior shirt of grid class, is particular about the striped of cut-parts that is sewed together, grid is mutually aligned, with satisfied visitor Family esthetic requirement, raising value of the product.But, owing to labor cost rises, weaving clothing competition among enterprises pressure is big, " work Method " to reasons such as the restrictions of workman overtime, weaving garment making industry needs a kind of new productivity to increase the performance of enterprises.Mesh Before, the weaving clothing enterprise of more domestic advanced persons has begun to explore full-automatic clothing technology, it is desirable to robot with The technology such as machine vision realize few artificial even unmanned clothing automatically.Therefore, research clothing cut-parts automatic sewing has great meaning Justice, and clothing cut-parts based on machine vision are its important link to bar to horizontal method.

The most common to horizontal document, method to bar about clothing cut-parts in image processing field research at present.

Summary of the invention

It is an object of the invention to provide a kind of clothing cut-parts based on images match method to horizontal stripe, the method is not required to Will be manually to horizontal stripe.

The technical solution adopted in the present invention is: a kind of clothing cut-parts based on images match method to horizontal stripe, specifically Implement according to following steps:

Step 1, the cut-parts sewing up needs carry out image acquisition；

Step 2, the image collected is carried out image gray processing；

Step 3, image to gray processing carry out Threshold segmentation, it is thus achieved that several cut-parts bianry images；

Step 4, to needing the every pair of bianry image to horizontal stripe to use Hole filling algorithms to carry out holes filling, use morphology Closed operation eliminates the jagged edges in cut-parts every pair bianry image；

Step 5, every pair of cut-parts bianry image are made and computing with the artwork of image acquisition in step 1, it is thus achieved that every pair is partitioned into Cut-parts image；

Step 6, the cut-parts image being partitioned into every pair carry out cut-parts positive direction correction respectively；

Step 7, choose every pair of region that wherein width cut-parts are waited in sewing as template；

Step 8, mate another width cut-parts region to be sewed, it is thus achieved that peak response position；

Step 9, calculate cut-parts displacement according to peak response position and template original position, move sanction according to displacement Sheet, completes cut-parts to horizontal stripe.

The feature of the present invention also resides in,

Step 1, particularly as follows: the cut-parts sewed up will be needed to be positioned on solid background plate, uses the pixel phase more than 5,000,000 Machine gathers image.

Step 2 carries out image gray processing according to formula (1),

Y=0.299R+0.587G+0.114B (1)

In formula (1), Y is brightness, and R, G, B are respectively the component that coloured image is red, green, blue.

Step 3 uses threshold binary image method that the image of gray processing is carried out Threshold segmentation,

Threshold calculations is carried out according to formula (2),

$Z^2\left\{\begin{array}{cc}255& f(x,y)\&GreaterEqual;T\\ 0& f(x,y)<T\end{array}\right.---\left(2\right)$

In formula (2), Z²Representing the bianry image after threshold operation, (x, y) represents original image pixels value to f, and T represents institute If threshold value, the span of T is 160-180.

Step 4 closing operation of mathematical morphology is processed as expanding image, corroding, and the area pixel size that burn into expands is about Being 20 × 16, burn into expansion structure element pixel size is 11 × 11.

Step 6 carries out cut-parts positive direction correction and specifically implements according to following steps the cut-parts image being partitioned into；

Step 6.1, the cut-parts image Q being partitioned into is solved minimum enclosed rectangle C, obtain minimum enclosed rectangle central point C_{(x, y)}；

Step 6.2, the minimum enclosed rectangle long side direction vector is made with horizontal line angle to beWith C_{(x, y)}For the center of circle by cut-parts Rotate counterclockwiseAgain to image Q₁Make minimum enclosed rectangle C₁, try to achieve minimum enclosed rectangle C₁Rectangular centre C_{(x1, y1)}And rectangle length of side L₁With wide L₂, with A width of L is made for rectangular centre₂, a length ofRectangle C₂, ask C₁And C₂Cut-parts size Area in two rectangles₁And Area₂, understand according to the observation, before and after just putting, above width, area is less than Area under, if Area₁≤Area₂, then cut-parts direction is positive direction；If Area₁＞ Area₂, then with C_(x1,y1)For the center of circle, will cut out Sheet rotates 180 °, finally gives cut-parts positive direction position.

Step 7 chooses every pair of region that wherein width cut-parts are waited in sewing: choose every pair wherein One width cut-parts wait that the zone level direction in sewing comprises 1 to 2 minimum cells, and vertical direction comprises 3 to 6 minimum lattice Subelement, stencil-chosen region treats sewing part in cut-parts, and comprises part non-cut-parts region, and search graph is to be matched as choosing / 6th size area of cut-parts, comprise sewing texure.

Step 8 utilizes Cross Correlation Matching algorithm based on gray scale, mates another width cut-parts region to be sewed, it is thus achieved that maximum sound Answer position, particularly as follows:

Size M × N image f (x, y) in move the subimage w of size J × K point by point (x y), make initial point and the point of w (x, y) overlaps, and calculates the sum of products of the image-region respective pixel covered by w in w with f, using this result of calculation as relevant Image (x, response c y) put (and x, y), with cross find peak response determine optimal matched position, formula such as formula (3) institute Show:

$c(x,y)=\frac{\underset{s=0}{\overset{K}{\&Sigma;}}\underset{t=0}{\overset{J}{\&Sigma;}}w(s,t)f(x+s,y+t)}{\sqrt{\underset{s=0}{\overset{K}{\&Sigma;}}\underset{t=0}{\overset{J}{\&Sigma;}}{w}^2(s,t)}\&times;\sqrt{\underset{s=0}{\overset{K}{\&Sigma;}}\underset{t=0}{\overset{J}{\&Sigma;}}{f}^2(x+s,y+t)}}---\left(3\right)$

In formula (3), (x, y) is cross-correlation calculation response value to c, and its value is between 0 to 1；(x y) is matching template to w；f (x y) is searched object；

(x, y) matrix, (x, y) in matrix, maximum is c (xMax, yMax) to c to calculate c by formula (3).

Step 9, particularly as follows: position in former cut-parts, the stencil-chosen region is [x, y], is rung according to the maximum that step 8 obtains Answer position c (xMax, yMax), then, be cut out template those width cut-parts vertically move distance for d=yMax-y, according to Vertically move distance and move cut-parts for d, complete cut-parts to horizontal stripe.

The invention has the beneficial effects as follows: a kind of clothing cut-parts based on images match method to horizontal stripe, it is not necessary to hands Work mode is to horizontal stripe, and the method uses image processing techniques to calculate the anglec of rotation and the displacement needing the cut-parts to horizontal stripe, Thus move for mechanical hand, cut-parts offer foundation is provided.

Accompanying drawing explanation

Fig. 1 is a kind of based on images match the clothing cut-parts using the present invention flow charts to the method for horizontal stripe.

Detailed description of the invention

The present invention is described in detail with detailed description of the invention below in conjunction with the accompanying drawings.

The invention provides a kind of clothing cut-parts based on images match method to horizontal stripe.

A kind of clothing cut-parts based on images match method to horizontal stripe, as it is shown in figure 1, specifically real according to following steps Execute:

Step 1, the cut-parts sewing up needs carry out image acquisition；

Step 1, particularly as follows: the cut-parts sewed up will be needed to be positioned on solid background plate, uses the pixel phase more than 5,000,000 Machine gathers image；

Step 2, the image collected is carried out image gray processing；Due to collected by camera to image be RGB image, directly Process data volume is big, and cut-parts segmentation is less demanding to color, therefore coloured image is carried out gray processing process,

Step 2 carries out image gray processing according to formula (1),

Y=0.299R+0.587G+0.114B (1)

In formula (1), Y is brightness, and R, G, B are respectively the component that coloured image is red, green, blue；

Step 3, image to gray processing carry out Threshold segmentation, it is thus achieved that several cut-parts bianry images；

Image segmentation is the dividing processing by acquired original image carries out certain mode, in order to carry from its result Get the process of some feature (such as profile, region etc.) of image, owing to the image that collects comprising the front width of cut-parts simultaneously Back panel, therefore needs to extract front and back's width respectively, is divided into a pair cut-parts, and striped alignment between a pair cut-parts, for adopting The image that collection arrives, uses threshold Image Segmentation method；

Step 3 uses threshold binary image method that the image of gray processing is carried out Threshold segmentation,

Threshold calculations is carried out according to formula (2),

$Z^2\left\{\begin{array}{cc}255& f(x,y)\&GreaterEqual;T\\ 0& f(x,y)<T\end{array}\right.---\left(2\right)$

In formula (2), Z²Representing the bianry image after threshold operation, (x, y) represents original image pixels value to f, and T represents institute If threshold value, the span of T is 160-180；

Step 4, to needing the every pair of bianry image to horizontal stripe to use Hole filling algorithms to carry out holes filling, use morphology Closed operation eliminates the jagged edges in cut-parts every pair bianry image；

There is hole in the bianry image being partitioned into due to reasons such as object complexity, these holes have impact on the complete of image object Whole property, simple Morphological scale-space cannot fill up these holes completely, and Hole filling algorithms can quickly be filled up in enclosing region Hole not of uniform size；

In step 4, closing operation of mathematical morphology is processed as expanding image, corroding, the area pixel size that burn into expands Being about 20 × 16, burn into expansion structure element pixel size is 11 × 11；

Hole filling algorithms can effectively eliminate inner void, but yet suffers from zigzag at boundary member, and i.e. part is on limit Half hole of edge is not padded；Expand image, corrode, i.e. closing operation of mathematical morphology processes, it is possible to fill edge disappearance Part, eliminates jagged edges, is conducive to extracting cut-parts shape completely.Expand, corrosion can reduce from boundary direction, increase Image size, the region that burn into expands depends on the size of structural element, the sawtooth sunk edge area size that threshold value produces, Its pixel size is about 20 × 16, selects 11 × 11 rectangular configuration elements expansions, corrosive effect preferable；

Step 5, every pair of cut-parts bianry image are made and computing with the artwork of image acquisition in step 1, it is thus achieved that every pair is partitioned into Cut-parts image；

Step 6, the cut-parts image being partitioned into every pair carry out cut-parts positive direction correction respectively；

The cut-parts that collection is arbitrarily put, as input picture, cause direction different, are difficult to obtain template and mate object, because of This, need to be corrected cut-parts；

Step 6 carries out cut-parts positive direction correction and specifically implements according to following steps the cut-parts image being partitioned into；

Step 6.1, the cut-parts image Q being partitioned into is solved minimum enclosed rectangle C, obtain minimum enclosed rectangle central point C_{(x, y)}；

Step 6.2, the minimum enclosed rectangle long side direction vector is made with horizontal line angle to beWith C_{(x, y)}For the center of circle by cut-parts Rotate counterclockwiseAgain to image Q₁Make minimum enclosed rectangle C₁, try to achieve minimum enclosed rectangle C¹Rectangular centre C_{(x1, y1)}And rectangle length of side L₁With wide L₂, with A width of L is made for rectangular centre₂, a length ofRectangle C₂, ask C₁And C₂Cut-parts size Area in two rectangles₁And Area₂, understand according to the observation, before and after just putting, above width, area is less than Area under, if Area₁≤Area₂, then cut-parts direction is positive direction；If Area₁＞ Area₂, then with C_(x1,y1)For the center of circle, will cut out Sheet rotates 180 °, finally gives cut-parts positive direction position；

Step 7, choose every pair of region that wherein width cut-parts are waited in sewing as template；

Cut-parts texture occurs in whole cut-parts repeatedly, arbitrarily intercepts template and matched position can be caused to offset, every a pair sanction Sheet seam area bottom curve radian is similar, can regard near symmetrical as, and this arcuate portion can be unique matching area, with Time, the reduction of template and object search area is conducive to computer disposal speed to improve,

Step 7 chooses every pair of region that wherein width cut-parts are waited in sewing: choose every pair wherein One width cut-parts wait that the zone level direction in sewing comprises 1 to 2 minimum cells, and vertical direction comprises 3 to 6 minimum lattice Subelement, stencil-chosen region treats sewing part in cut-parts, and comprises part non-cut-parts region, and search graph is to be matched as choosing / 6th size area of cut-parts, comprise sewing texure, and error deviation is minimum；

Step 8, mate another width cut-parts region to be sewed, it is thus achieved that peak response position；

Step 8 utilizes Cross Correlation Matching algorithm based on gray scale, mates another width cut-parts region to be sewed, it is thus achieved that maximum sound Answer position, particularly as follows:

Size M × N image f (x, y) in move the subimage w of size J × K point by point (x y), make initial point and the point of w (x, y) overlaps, and calculates the sum of products of the image-region respective pixel covered by w in w with f, using this result of calculation as relevant Image (x, response c y) put (and x, y), with cross find peak response determine optimal matched position, formula such as formula (3) institute Show:

$c(x,y)=\frac{\underset{s=0}{\overset{K}{\&Sigma;}}\underset{t=0}{\overset{J}{\&Sigma;}}w(s,t)f(x+s,y+t)}{\sqrt{\underset{s=0}{\overset{K}{\&Sigma;}}\underset{t=0}{\overset{J}{\&Sigma;}}{w}^2(s,t)}\&times;\sqrt{\underset{s=0}{\overset{K}{\&Sigma;}}\underset{t=0}{\overset{J}{\&Sigma;}}{f}^2(x+s,y+t)}}---\left(3\right)$

In formula (3), (x, y) is cross-correlation calculation response value to c, and its value is between 0 to 1；(x y) is matching template to w；f (x y) is searched object；

(x, y) matrix, (x, y) in matrix, maximum is c (xMax, yMax) to c to calculate c by formula (3)；

Step 9, calculate cut-parts displacement according to peak response position and template original position, move sanction according to displacement Sheet, completes cut-parts to horizontal stripe；

Step 9, particularly as follows: position in former cut-parts, the stencil-chosen region is [x, y], is rung according to the maximum that step 8 obtains Answer position c (xMax, yMax), then, be cut out template those width cut-parts vertically move distance for d=yMax-y, according to Vertically moving distance and move cut-parts for d, horizontal displacement is sewed up mechanical requirements according to reality and is determined, completes cut-parts to horizontal stripe.

The cut-parts anglec of rotation asked according to above-mentioned steps and displacement, these parameter point two ways apply to industry On, one: using scaling method, pixel unit is converted into actual physics coordinate, mechanical hand moves cut-parts according to physical distance； Its two: mobile cut-parts in the range of pixel definition, until cut-parts error distance is less than specified pixel error.

A kind of clothing cut-parts based on images match method to horizontal stripe, it is not necessary to manually to horizontal stripe, the method Use image processing techniques to calculate the anglec of rotation and the displacement needing the cut-parts to horizontal stripe, thus move for mechanical hand, stitch Sanction sheet and foundation is provided.

Claims (9)

1. clothing cut-parts based on the images match method to horizontal stripe, it is characterised in that implement according to following steps:

Step 1, the cut-parts sewing up needs carry out image acquisition；

Step 2, the image collected is carried out image gray processing；

Step 3, image to gray processing carry out Threshold segmentation, it is thus achieved that several cut-parts bianry images；

Step 4, to needing the every pair of bianry image to horizontal stripe to use Hole filling algorithms to carry out holes filling, close fortune with morphology Calculate the jagged edges eliminated in cut-parts every pair bianry image；

Step 5, every pair of cut-parts bianry image are made and computing with the artwork of image acquisition in step 1, it is thus achieved that every pair of sanction being partitioned into Picture；

Step 6, the cut-parts image being partitioned into every pair carry out cut-parts positive direction correction respectively；

Step 7, choose every pair of region that wherein width cut-parts are waited in sewing as template；

Step 8, mate another width cut-parts region to be sewed, it is thus achieved that peak response position；

Step 9, calculate cut-parts displacement according to peak response position and template original position, move cut-parts according to displacement, Complete cut-parts to horizontal stripe.

A kind of clothing cut-parts based on images match method to horizontal stripe, it is characterised in that described Step 1, particularly as follows: the cut-parts sewed up will be needed to be positioned on solid background plate, uses the pixel collected by camera figure more than 5,000,000 Picture.

A kind of clothing cut-parts based on images match method to horizontal stripe, it is characterised in that described Step 2 carries out image gray processing according to formula (1),

Y=0.299R+0.587G+0.114B (1)

In formula (1), Y is brightness, and R, G, B are respectively the component that coloured image is red, green, blue.

A kind of clothing cut-parts based on images match method to horizontal stripe, it is characterised in that

Described step 3 uses threshold binary image method that the image of gray processing is carried out Threshold segmentation,

Threshold calculations is carried out according to formula (2),

$Z^2=\left\{\begin{array}{cc}255& f(x,y)\&GreaterEqual;T\\ 0& f(x,y)<T\end{array}\right.---\left(2\right)$

In formula (2), Z²Representing the bianry image after threshold operation, (x, y) represents original image pixels value to f, and T represents set threshold Value, the span of T is 160-180.

A kind of clothing cut-parts based on images match method to horizontal stripe, it is characterised in that described Step 4 closing operation of mathematical morphology is processed as expanding image, corroding, and the area pixel size that burn into expands is about 20 × 16, Burn into expansion structure element pixel size is 11 × 11.

A kind of clothing cut-parts based on images match method to horizontal stripe, it is characterised in that described Step 6 carries out cut-parts positive direction correction and specifically implements according to following steps the cut-parts image being partitioned into；

Step 6.1, the cut-parts image Q being partitioned into is solved minimum enclosed rectangle C, obtain minimum enclosed rectangle central point C_{(x, y)}；

Step 6.2, the minimum enclosed rectangle long side direction vector is made with horizontal line angle to beWith C_{(x, y)}For the center of circle by the cut-parts inverse time Pin rotatesAgain to image Q₁Make minimum enclosed rectangle C₁, try to achieve minimum enclosed rectangle C₁Rectangular centre C_{(x1, y1)} And rectangle length of side L₁With wide L₂, with A width of L is made for rectangular centre₂, a length ofRectangle C₂, seek C₁And C₂ Cut-parts size Area in two rectangles₁And Area₂, understand according to the observation, before and after just putting, above width, area is less than lower aspect Long-pending, if Area₁≤Area₂, then cut-parts direction is positive direction；If Area₁＞ Area₂, then with C_(x1,y1)For the center of circle, cut-parts are rotated 180 °, finally give cut-parts positive direction position.

A kind of clothing cut-parts based on images match method to horizontal stripe, it is characterised in that described Step 7 chooses every pair of region that wherein width cut-parts are waited in sewing: choose every pair of wherein width cut-parts Waiting that the zone level direction in sewing comprises 1 to 2 minimum cells, vertical direction comprises 3 to 6 minimum cells, Sewing part is treated in cut-parts in stencil-chosen region, and comprises part non-cut-parts region, and cut-parts to be matched chosen by search graph picture / 6th size area, comprise sewing texure.

A kind of clothing cut-parts based on images match method to horizontal stripe, it is characterised in that described Step 8 utilizes cross-correlation method, mates another width cut-parts region to be sewed, it is thus achieved that peak response position, particularly as follows:

Size M × N image f (x, y) in move point by point size J × K subimage w (x, y), make the initial point of w and point (x, y) Overlap, calculate the sum of products of the image-region respective pixel covered by w in w with f, this result of calculation is existed as associated picture (x, response c y) put (x, y), determines optimal matched position, shown in formula such as formula (3) with crossing searching peak response:

$c(x,y)=\frac{\underset{s=0}{\overset{K}{\&Sigma;}}\underset{t=0}{\overset{J}{\&Sigma;}}w(s,t)f(x+s,y+t)}{\sqrt{\underset{s=0}{\overset{K}{\&Sigma;}}\underset{t=0}{\overset{J}{\&Sigma;}}{w}^2(s,t)}\&times;\sqrt{\underset{s=0}{\overset{K}{\&Sigma;}}\underset{t=0}{\overset{J}{\&Sigma;}}{f}^2(x+s,y+t)}}---\left(3\right)$

Wherein (x, y) is cross-correlation calculation response value to c, and its value is between 0 to 1；(x y) is matching template to w；(x, y) for be searched for f Rope object；

(x, y) matrix, (x, y) in matrix, maximum is c (xMax, yMax) to c to calculate c by formula (3).

A kind of clothing cut-parts based on images match method to horizontal stripe, it is characterised in that described Step 9 is particularly as follows: position in former cut-parts, the stencil-chosen region is [x, y], according to the peak response position c of step 8 acquisition (xMax, yMax), then, be cut out template those width cut-parts vertically move distance for d=yMax-y, according to vertically moving Distance moves cut-parts for d, completes cut-parts to horizontal stripe.